Experiments.

(1)

Using Panlab Federation Mechanisms and Infrastructure for Cloud

Experiments

(2)

Panlab overview

 Initial federation and Panlab concepts started in 2006/2007 with the Panlab SSA (FP6)

 The FP7 IP Pan-European Laboratory Infrastructure Implementation (PII) started in

2008 and will run until Q4 2010.

 PII implements the Panlab concept of cross-domain, cross-layer resource federation

 Target is to create a large scale testing and experimentation facility by integrating

existing and emerging testbeds

A hi l d i i i d d h h f d i

 Achieve scale and maintain independence through federation

– that enables end-to-end interoperability testing of platforms, networks and services – that helps reducing the risks and costs of large-scale network infrastructure testing

(3)

The high level Panlab architecture

 Various testbeds provide resources that are controlled via a federation framework and user frontend (Teagle) and user frontend (Teagle)

 With the help of Teagle resources can be combined and configured for specific a d co gu ed o spec c setups

 Panlab aims at satisfying broad R&D requirements by providing a generic resource control framework for

distributed heterogeneous resources

resources

 Federation with other

frameworks is currently under investigationg

(4)

Functions & Interfaces

T R i t & R it U1 Customer Domain End User Domain T EAGLE

Registry & Repository Creation Environment Request Processor Orchestration Engine Web Portal U2 T1 U3 RA Domain Manager RA RA RA Domain Manager RA RA Domai IGW T2 T2 I2 RA RA RA m ain B IGW T2 RA RA RA T2 A1 n A I1 B1 Do m

RA: resource adapter

IGW: interconnection gateway A1: resource in domain A B1: resource in domain B

(5)

Demo video



A screen cast demonstration video was shown during the

presentation. The video can soon be found here:

http://www.fire-teagle.org/tutorials.jsp

(6)

The Demo

www.openimscore.org

www.fire-teagle.org

A custom testbed is designed with Teagle making use of infrastructure resources (hardware, virtual machines) and service layer resources (Open Source IMS core)

(7)

The Experiment

www.openimscore.org

h

stress the core …

… and break it

Load generator

… and break it

Load generator

www.sipnuke.org

* 50 IMS USER IDs

* 5 REGISTER per secondp

* increasing from 50 to 400 CALLs per second, looping through all users * UNREGISTER all users

(8)

Second experiment execution

modify testbed

www.fire-teagle.org

(9)

Modify the custom testbed

virtual machine

hosting the core

384 MB RAM 1 CPU

www.openimscore.org

compute resource

+ configuration of the new P-CSCF and S-CSCF that have been moved onto a cloud resource

network resource storage

Cloud Resource

+ re-configuration of the old core to deal with the new instances + re configuration of DNS server etc

g resource

(10)

Behind the scenes

The Teagle tool helps fi i th t tb d reconfiguring the testbed and deploying new resource instances.

From the design

environment requests are send to the repository to p y update resource

configurations and reservations.

The topology-oriented

testbed design is send to the Orchestration Engine by

Orchestration Engine by means of a XML document.

(11)

The Orchestration Engine transforms the design environment output into an executable script and executes it (resolving dependencies, etc.)

(12)

RA

From the script executor provided by the Orchestration From the script executor provided by the Orchestration Engine, generic REST requests (CRUD) are send via interface T1 to those domain managers (PTM) that are responsible for involved resources.

The PTM passes the generic CRUD (create, read, update, delete) requests to the resource adaptors (RA) that control the involved resources. Here, the generic commands are translated into resources specific requests and actions translated into resources specific requests and actions.

(13)

Some screenshots from the demo screen cast

Initial testbed design using the Teagle

creation environment. Th d tt d li fl t The dotted lines reflect a containment

relationship (e.g. the ICSCF is hosted by the y XENnode). The solid lines represent a

(14)

Some screenshots from the demo screen cast

Execution of the first

experiment. The CPU load of the small XEN reaches of the small XEN reaches 100%, additionally the machine runs out of memory.

Therefore, delays and high call error rates are observed.

(15)

Some screenshots from the demo screen cast

The IMS Core logs show out of memory errors. At this stage the core is

this stage the core is unstable and high delays and error rates are

(16)

Some screenshots from the demo screen cast

The testbed after the re The testbed after the re-design.

The IMS Core services PCSCF and SCSCF have PCSCF and SCSCF have been moved onto the cloud computing resource and are now running “in the

l d” cloud”.

The resources that are in

state: new are additionally

booked into the testbed by booked into the testbed by sending created and update commands to the involved domains.

(17)

Some screenshots from the demo screen cast

As expected the metrics As expected the metrics show a much better performance now. 400 calls per second could be p handled easily after

modifying the testbed setup.

N d hi h

No errors and no higher delays than 0.2ms.

The CPU load on the computing resource (4 computing resource (4 CPUs, 4GB mem) offered by the cloud stays below 20%.

(18)

Thank You & Contact

Sebastian Wahle

Dipl -Ing Dipl.-Ing.

Next Generation Network Infrastructures

Fraunhofer Institute for Open Communication Systems Kaiserin-Augusta-Allee 31 | 10589 Berlin | Germany Phone +49 30 3463 -7365 | Fax -8000|

[email protected] www.fokus.fraunhofer.de

(19)

Further Information

Some Links & Publications

 Panlab main website: http://panlab.org/

 Teagle website: http://www.fire-teagle.org/

 VCT Tool information & tutorials: http://www.fire-teagle.org/tutorials.jsp

 Last year’s FIA book:

Anastasius Gavras, Halid Hrasnica, Sebastian Wahle, David Lozano, Denis Mischler, and Spyros Denazis. Towards the Future Internet - A European Research Perspective, chapter Control of Resources in Pan-European Testbed

Federation, pages 67 - 78. IOS Press, May 2009. ISBN 978-1-60750-007-0.ede at o , pages 6 8 OS ess, ay 009 S 9 8 60 50 00 0

 Forthcoming:

Konrad Campowsky, Anastasius Gavras, Bogdan Harjoc, Thomas Magedanz, and Sebastian Wahle. Pan-European Testbed and Experimental Facility Federation – Architecture Refinement and Implementation. Inderscience

International Journal of Communication Networks and Distributed Systems (IJCNDS), Special Issue: Recent Advances in Test-bed Driven Networking Research. Accepted for publication in 2010.

Konrad Campowsky, Thomas Magedanz, and Sebastian Wahle. Resource Management in Large Scale Experimental Facilities: Technical Approach to Federate Panlab and PlanetLab. In 12th IEEE/IFIP Network Operations and

Management Symposium (NOMS 2010). IEEE/IFIP, 2010. Accepted for publication in 2010.